AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Apate sends AI agents undercover to get scammed so they can gather fraud intelligence

Collected Oct 14, 2026

Apate.ai, an Australian startup founded in 2024 on research from Macquarie University in Sydney, runs a fleet of thousands of AI agents whose job is to be scammed. The agents answer scam calls, reply to suspicious messages and emails, follow phishing links and join social media groups, posing as victims of varying ages, genders, accents and languages. The company says the fleet has consumed 17 years of scammers' time across 2.5 million conversations, and that the point is to extract details normally learned only after someone has already been defrauded: bank accounts, crypto wallets, phone numbers, malicious websites, the brands being impersonated, and other specifics of the operations.

Those findings are passed to Apate's customers, which include banks, telecom operators, government agencies and utilities. Customers also get an insights portal with intelligence and analytics about scams hitting their end users, employees, brands and products. They can task agents with particular objectives, such as obtaining a bank account number or a crypto wallet address, and the text agents can scan forums, investment groups and social platforms for suspicious activity before starting a fabricated conversation.

Apate is one of the finalists in this year's Startup Battlefield 200 at TechCrunch Disrupt, held at Moscone West in San Francisco through Thursday, October 15, where it will pitch a panel of venture capital judges. Named customers include Commonwealth Bank of Australia and TPG Telecom, along with other unnamed utilities, telcos and government agencies. According to co-founder and chief commercial officer Brad Joffe, TPG has diverted nearly 590,000 calls to Apate's agents, Commonwealth has recorded a 76% drop in publicly reported scam losses, and an unnamed Australian bank used its outbound agents to disrupt 22 scam campaigns.

On the technical side, Joffe said Apate combines frontier models with its own fine-tuned models, trained on the conversations its agents have had with scammers plus the earlier Macquarie research. That research, by founder and CEO Dali Kaafar and colleagues, applied machine learning and natural language processing to the content of scam calls and to better detection and prevention methods.

The origin story is a real scam call. Several years ago, Kaafar was at a family picnic when a call came from an unknown number. He realized it was a scammer and played dumb, keeping the caller talking for 44 minutes. The realization that those were 44 minutes the scammer was not spending on a real victim led him to ask whether the process could be automated and made to yield useful information.

The broader context explains why investors are interested. The FBI received more than 22,000 complaints of AI-related crimes in 2025, with roughly $893 million in losses, according to the article. More than $632 million of that came from investment scams, where the FBI says fraudsters use AI to run conversations and deepfake the voices and appearances of celebrities, executives and trusted figures.

Joffe framed Apate's approach as a departure from post-hoc defense. Most of the industry works after the fact, he argued: telcos block numbers and scammers get new ones, banks watch transactions but only after a scammer has reached the customer. Apate's pitch is to go to the scammer first and gather information before anyone is harmed. The agents talk to tens of thousands of scammers daily across calls, messages, email and the web, and the product is what emerges from those conversations.

Scam-baiting has long been an amateur pursuit, with people stringing fraudsters along for entertainment. Automating it changes the economics. An agent can hold many conversations at once, at negligible marginal cost, and can be pointed at specific intelligence targets. This suggests the value is less in the prank and more in the structured output: indicators that a bank can block, numbers a telco can act on, campaigns a security team can trace.

There are limits worth noting. Detection and blocking are adversarial and time-sensitive, so intelligence has a shelf life, and scammers who suspect a bot will change numbers, scripts and infrastructure. Agent-based engagement also raises questions about what data is collected during baiting and how third parties who appear in those conversations are handled, though the company describes its output as scammer-side details and cross-customer analytics rather than victim data. A likely trade-off is coverage versus depth: a fleet built to keep scammers talking optimizes for conversation length and coverage, while extracting account numbers and wallet addresses may require longer, more targeted engagements.

On funding, Apate raised $8.15 million in seed capital in August from Lobby Capital, OIF Ventures, Investible, Concept Ventures and Baobab Ventures. Joffe said the 20-person company has raised just under $10 million total and will spend the round on product and international expansion, including a London office and a larger U.S. presence. The company takes its name from the Greek goddess of deception, and Joffe noted that scammers swearing at the bots feature in its pitch materials; Apate treats profanity as a sign of frustration and celebrates it.

Why it matters: Banks, telcos, utilities and government anti-fraud teams are the immediate audience, since Apate sells the output of scammer conversations rather than software that blocks traffic in real time. For product and engineering teams, the model points to a different way of using agents, as long-running sensors that operate against adversaries rather than as workflow automation, which implies new work in identity fabrication, conversation steering, abuse safeguards and turning unstructured chats into reliable threat feeds. Whether this holds up depends on measurable outcomes like Commonwealth's reported drop in publicly reported losses, which Apate cites as a proxy for effectiveness.

Read at TechCrunch · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Most companies use AI agents to ship products faster or speed up workflows. Apate.ai sends them undercover to get scammed.